The accuracy of PER ratings

[SAGambler]

I've heard it said that PER really isn't an accurate statistic. However, this may be a chance to prove or disprove how it measures a teams chances.

Currently the Spurs have two players, Ginoboli and Duncan in the top 5. No other team has 2 players even in the top 10.

So, if the Spurs go on to win another championship, would this indicate that PER ratings do in fact give an indication of how well teams are playing?

Discuss

[sabar]

PER is an excellent measure of TANGIBLES, and hence usually captures how good a player is. But until things like touches, deflections, and intangibles can be quantified, it won't be the end-all statistic that Hollinger would like you to think it is.

[tav1]

I think the PER is generally very accurate given what it is: a per minute rating that is not too concerned with defense. I suspect Hollinger will continue to tweak the rating and that it will become more reliable in time. I would prefer it included deflections, +/-, charges taken, etc...

[travis2]

It's also only accurate for starters and regular subs...ie, those who play a significant number of minutes. It also isn't ready to be taken seriously until late in the season.

[tp2021]

All calculations begin with what is called unadjusted PER (uPER). The formula is:

\ uPER = \frac{1}{Min} *\left ( 3P + [(2/3)*AST] + [(2 - factor*(tmAST/tmFG))*FG] + [FT*0.5*(1 + (1 - (tmAST/tmFG)) + (2/3)*(tmAST/tmFG))] - [VOP*TO] - [VOP*DRBP*(FGA - FG)] - [VOP*0.44*(0.44 + (0.56*DRBP))*(FTA - FT)] + [VOP*(1 - DRBP)*(TRB - ORB)] + [VOP*DRBP*ORB] + [VOP*STL] + [VOP*DRBP*BLK] - [PF*((lgFT/lgPF) - 0.44*(lgFTA/lgPF)*VOP))] \right )

Where

* \ factor = (2/3) - [(0.5*(lgAST / lgFG)) / (2*(lgFG / lgFT))] ,
* \ VOP = [lgPTS / (lgFGA - lgORB + lgTO + 0.44*lgFTA)] ,
* \ DRBP = [(lgTRB - lgORB) / lgTRB] .

Once uPER is calculated, it must be adjusted for team pace and normalized to the league to become PER:

\ PER = [uPER*(lgPace/tmPace)] * (15/lguPER)

This final step takes away the advantage held by players whose teams play a fastbreak style (and therefore have more possessions and more opportunities to do things on offense), and then sets the league average to 15.00.

Also note that it is impossible to calculate PER (at least in the conventional manner described above) for NBA seasons prior to 1978, as the league did not keep track of turnovers before that year.




all that courtesy of wiki. just thought i'd put it for people who'd like to know the calculations.

[tp2021]

heres the link, its more clear on the actual page now that i look at it.


http://en.wikipedia.org/wiki/Player_Efficiency_Rating